How does cross-device advertising work: the complexity and prospects of technology development

Introduction

An increasing number of users are accessing the network through various devices. In this case, the interaction of the advertiser with the potential buyer occurs using a variety of advertising channels. Often, the device from which a person consumes content and defines this channel of interaction. The user may be interested in television advertising, or vice versa - get distracted during its display for communication in a social network on a mobile phone or personal computer. A potential buyer on the way from the first contact with a brand or product to the time of purchase can change more than one device, and at the same time it will not always be personal.



According to data from [1], 95% of Russians have a mobile phone, while only 80% use a laptop or personal computer. According to Google Russia [2], in 2014, 62% used mobile devices to search for information about products, and 39% of domestic users made a purchase from a smartphone at least once. It also notes that the path to purchase, for example, in the retail segment, which began with a search on a mobile device, ended with a purchase on the same device only in 3% of cases.

In turn, the specialists of Criteo predicted [3] that the increase in the number of purchases made in RuNet from mobile devices in 2016 will exceed 50%.
')

Fig. 1 - Owners of digital devices among the adult population of the Russian Federation according to data from [1].

The above determines the need to assess the effectiveness of advertising campaigns through all the channels used and on all the devices through which the advertiser interacts with the buyer.

Using the method of creating linked ads for various devices, including personalizing the promotional offer for users who have several sources of access to the network. The so-called related advertising has many English synonyms and is popularized under a lot of selling phrases, for example:

• Tapad : Personalize content across multiple screens.
• BlueCava : Reach high-value targets on any screen.
• Adelphic : Target people, not devices.
• Tactads / MediaMath : On the right devices.
• Drawbridge : Accurately scale desktop retargeting campaigns to mobile.

Today we will talk about how this works and what prevents the widespread adoption of this approach.

With the advent of an ecosystem for automated purchasing advertising, an advertiser can identify his user not only within a specific site, but also within all sites participating in an advertising campaign. This allows you to reliably assess the scope of an advertising campaign and adjust the frequency of displaying an ad for a unique user. However, in order to provide such a service, the advertising platform must do a careful job of comparing the set of user identifiers to one universal one, within which: the frequency of the show is set up, the classroom targeting; monitored the effectiveness of the advertising campaign.

Globally, we can distinguish the following tasks in one way or another related to the work of comparing many different user identifiers to one universal one:

• User identification on various SSP ( Supply / Sell Side platform );
• user authentication between browsers;
• User identification between browsers and applications on a mobile device;
• User identification between various devices - as the top of the evolutionary development of advertising technologies.

Fig. 2 - Illustrations for cross-identification of the user: (a) - User identification on different SSPs, (b) - User identification between browsers, (c) - User identification between browsers and applications on a mobile device.

User identification on different SSPs

One DSP ( Demand-Side Platform ) is always connected to more than one SSP, and one and the same user (observed through different SSP) even from one device can be considered as part of an advertising campaign as several different people. This does not allow to reliably observe the frequency of advertisements displayed for a unique user.

The Open RTB protocol [4] from the IAB for the User object provides two fields for identifying a user at the time of the request ( bid ) from the SSP to the DSP: id — user ID within the specific SSP and buyerid — the DSP user ID. In order for the SSP to transmit information about the buyerid, it is necessary to implement the technology of identifier matching on the SSP side, for example [5]. However, not all SSPs are often ready to store a very large match table on their side and spend technological resources on it.

In practice, it is often done differently - the SSP sends its user ID to the DSP side as part of the identifier matching process. In this case, the SSP is enough to fill in the id field in the User object and the DSP will independently understand for which user the potential display is offered. More rarely, two-way exchange of identifiers is used.

The careful work of the DSP in cross-identification of users received through various SSPs allows us to reliably observe the frequency of impressions required by the advertiser and to ensure greater coverage of the advertising campaign.

User identification in browsers and applications

It is not uncommon for users to identify a user as part of a single browser, this may be due to the fact that a user uses special browser extensions for anonymization purposes and setting a certain type of cookie is impossible (especially important for Safari on mobile devices). This immediately leads to the fact that advertising traffic from such users cannot be used even for the needs of classical retargeting and control over the frequency of displaying advertisements to such a user stops working. To solve the described problem and identify these users, fingerprint technology is usually used (for example, Panopticlick , [6]), a rather complete comparison of which various implementations is presented in [7]. It also shows that a user who uses certain anonymization technologies is the opposite - more vulnerable from the point of view of anonymity on the network.

When developing the Exebid.DCA DSP , we acted much easier by stopping experiments on the intentional identification of the user, which is obviously at the moment against this. We managed to isolate a segment of such users and rid them of advertising campaigns in which sustainable identification is important. In more simple cases, when you can not put the so-called. Third-party cookies can use the capabilities of modern browsers such as Local storage ( localStorage ) and postMessage [8].

For desktop devices, the situation is also quite common when a user uses two different browsers for different tasks. If you do not take special measures of identification, then such a user will be considered by the advertising platform as several different users who are not connected with each other. Here, for identification, browser-independent features are used, for example, from [9]: IP address, information about installed fonts, time zone, screen resolution, etc.

A separate problem is the comparison of user behavior within the mobile application and on the web site. In general, this problem does not have a solution, but to track the effectiveness of advertising in mobile applications, you can use the technology of “native click” ( Native Browser Click , for example [10]). In this case, it is possible to compare IDFA ( Identifier for Advertizer ) from the application with the user's cookie in the mobile device browser.

Choose on the phone and buy from a personal computer

The tendency of modern Internet buyers to get acquainted with advertising offers from tablets or smartphones, and later to buy goods from a personal computer and a senior analyst at Google Russia Stanislav Vidyaev. “Making a purchase or ordering a service from a small screen of a smartphone or tablet is more difficult than from a desktop,” he said at the Google Think Performance conference [11], leading the audience to the idea of ​​the inevitability of introducing cross-device tracking of unique users.

Some devices help others convert

The ability to connect the user's behavior patterns with a smartphone / tablet and PC and determine that this is the same person will greatly simplify the life of advertisers. Using the new technology, they will be able to push indecisive visitors to the site or mobile application to re-visit from another device and purchase (remarketing based on cross-device user ID), as well as save the budget without showing ads on the tablet to those who have already ignored it smartphone or desktop computer, and vice versa.

In addition to the introduction of the cross-device advertising model, Internet users themselves should remain. They will begin to receive only interesting personalized offers and get rid of the tedious flow of the same irrelevant advertising for them on all devices.

Although cross-device advertising has not yet had time to become an objective reality due to a number of technical difficulties, some companies have already managed to try out a new method and get the first results. So, the journalists of the American specialized publication Adweek [12] learned about the results of an advertising campaign of a luxury car manufacturer who used retargeting technology for users who viewed ads on three different devices - smartphones, tablets and personal computers. It turned out that the automaker increased conversion by 15% compared to contacts on a single device.

Approaches to cross-user identification

But how to determine that the smartphone, tablet and PC are the same owner? Google Analytics offers to identify unique users not only by tracking the cookies of browsers on a personal computer or id-mobile device, but also by identifying the customer of a particular store (this causes the cross-device matching technology to work only for the registered audience of the site). The same user, sitting with different devices, the system recognizes after authorization. According to [11] Stanislav Vidyaev of Google, Russia, it helps advertisers to track the user's path to purchase and avoid erroneous judgments such as “advertising for tablets is ineffective because all transactions are made using personal computers . ”

Such a deterministic approach ( deterministic tracking ) is actually based on matching user logins on Internet systems, it can be not only a specific target site-store, but also social networks and even browsers, for example, synchronization of settings and browsing history in Google Chrome browser between different devices available from version 18, 2012 [13]. Thus, this approach could be implemented and popularized much earlier.

Any platform and publishers that collect user credentials can use deterministic tracking. However, not for all users, the path to the purchase consists of two simple steps: viewing an offer from a mobile device or tablet, placing an order through a personal computer; and not everyone wants to register on the site.

The second, and more complex approach to cross-device matching is probabilistic tracking ( probabilistic tracking ), which involves the use of probabilistic algorithms to analyze user behavior on various devices. Large technology companies, such as BlueCava, Adelphic, Tapad and Drawbridge or Russian DCA, collect data on multiple cookies, analyze similar patterns of search engine usage and use special tests to determine the connection of different devices with the same user profile.

The most obvious approach to the implementation of cross-device technology is to consider the whole set of identifiers of all devices of all users as a graph whose vertices are these identifiers, and the weighted connection of two identifiers with an edge means that these identifiers belong to one person with a certain probability, for example, [14]. This approach allows using known methods of searching clusters in social networks and identifying the same user on different devices.

According to Drawbridge, probabilistic tracking can guarantee 97.3% accuracy [15]. In 2013, Expedia (a site for online booking of tickets and hotels) conducted tests of this system, during which an announcement was offered on the mobile phones of the website with a proposal to install Expedia [16]. According to the test results, an increase in conversion rate by whole orders was noted.

Conclusion and development prospects

Even complex technologies of cross-device advertising, which allow connecting several devices connected to a network with a user-buyer, this is not the limit of the development of digital tools. Cross-device advertising is the first step towards perfect cross-channel convergence, which involves broadcasting the same ad form to interested parties through several channels at once - Internet sites, email distribution, television, radio, social networks, call centers companies. For example, a viewer who saw a commercial and decided to find out more about the product by calling the customer service specialist may send an email with an offer to purchase goods at a discount or get a bonus for the purchase.

The introduction of an efficiently working cross-channel model will become a marketing revolution, but specialists will be able to work on its development closely only when a cross-device model works without flaws. Now we are seeing difficulties with the convergence of devices, even within a single channel - the Internet. A user logging on with two browsers is still perceived by the system as two different people. Moreover, on mobile devices, the user inside the application and in the browser (of which there can be several) is counted as several users. Therefore, the development of IDFA matching technology, which is used to identify a user within applications, with a default browser of a mobile device is an actual popular task.

In Russia, additional difficulties are added to this: in many families, only smartphones are “personal”, and several family members often use tablets and PCs at once, and advertising offers in this case may not reach their destination.

Despite the apparent difficulties, it is obvious - for the cross-device advertising future. That is why business needs to develop methods for collecting and processing information about potential customers and their needs.